R, xgboost: label must be in [0,1] for logistic regression

问题

I am getting an error saying that my label must be in [0, 1]:

> system.time(xgb <- xgboost(params  = param,
+                            data    = dtrain,
+                            label   = as.numeric(train.label),
+                            nrounds = 500,
+                            print_every_n = 100,
+                            verbose = 1))
Error in xgb.iter.update(bst$handle, dtrain, iteration - 1, obj) : 
  [10:39:29] amalgamation/../src/objective/regression_obj.cc:108: label must be in [0,1] for logistic regression
Timing stopped at: 0.11 0 0.11

However, my label is in [0, 1]:

> train.label
   [1] 1 1 1 1 1 0 0 1 1 0 0 1 1 1 1 0 1 1 1 1 1 0 1 1 . . .
 [ reached getOption("max.print") -- omitted 38907 entries ]
Levels: 0 1

I also tried converting to a numeric data type without luck.

回答1:

The problem is that your train.label is a factor, so your code

as.numeric(train.label)

will produce a vector containing 1's and 2's. You want the 0's and 1's as the values, so you need to use.

as.numeric(as.character(train.label))

来源：https://stackoverflow.com/questions/48999589/r-xgboost-label-must-be-in-0-1-for-logistic-regression